Guide the probability to scatter in or transmit through the volume.
Only applied for primary rays.
Co-authored-by: Brecht Van Lommel <brecht@blender.org>
The distance sampling is mostly based on weighted delta tracking from
[Monte Carlo Methods for Volumetric Light Transport Simulation]
(http://iliyan.com/publications/VolumeSTAR/VolumeSTAR_EG2018.pdf).
The recursive Monte Carlo estimation of the Radiative Transfer Equation is
\[\langle L \rangle=\frac{\bar T(x\rightarrow y)}{\bar p(x\rightarrow
y)}(L_e+\sigma_s L_s + \sigma_n L).\]
where \(\bar T(x\rightarrow y) = e^{-\bar\sigma\Vert x-y\Vert}\) is the
majorant transmittance between points \(x\) and \(y\), \(p(x\rightarrow
y) = \bar\sigma e^{-\bar\sigma\Vert x-y\Vert}\) is the probability of
sampling point \(y\) from point \(x\) following exponential
distribution.
At each recursive step, we randomly pick one of the two events
proportional to their weights:
* If \(\xi < \frac{\sigma_s}{\sigma_s+\vert\sigma_n\vert}\), we sample
scatter event and evaluate \(L_s\).
* Otherwise, no real collision happens and we continue the recursive
process.
The emission \(L_e\) is evaluated at each step.
This also removes some unused volume settings from the UI:
* "Max Steps" is removed, because the step size is automatically specified
by the volume octree. There is a hard-coded threshold `VOLUME_MAX_STEPS`
to prevent numerical issues.
* "Homogeneous" is automatically detected during density evaluation
An option "Unbiased" is added to the UI. When enabled, densities above
the majorant are clamped.
Due to numerical issues this was creating many wrong self-overlapping.
It was necessary for skipping empty regions, but not any more with the
volume Octree approach
Since we sample the same light for distance sampling and equiangular
sampling, the sample is invalid anyway, so just avoid sampling direct
light for distance sampling too.
This follows the other CMake "modernization" commits, this time for
`bf_intern_openvdb` and the OpenVDB dependency itself.
The difference with this one is that `intern/openvdb` becomes an
"optional" dependency itself. This is because downstream consumers often
want to include this dependency rather than openvdb directly, so this
target must also be optional. Optional, in this case, means the target
always exists but may be entirely empty.
Summary
- If you are using BKE APIs to access openvdb features, then use the
`bf::blenkernel` target
- If you are only using `intern/openvdb` APIs then use the
`bf::intern::optional::openvdb` target (rare)
- For all other cases, use the `bf::dependencies::optional::openvdb`
target (rare)
context: https://devtalk.blender.org/t/cmake-cleanup/30260
Pull Request: https://projects.blender.org/blender/blender/pulls/137071
Use RAII for `GHOST_WindowClass`.
For the main Blender executable, only one `GHOST_System` is ever
created. But when running tests, this can happen multiple times (eg,
GPUOpenGLTest and GPUOpenGLWorkaroundsTest).
This results in `GHOST_SystemWin32` trying to register
`GHOST_WindowClass` multiple times, so `RegisterClassW` fails with
`ERROR_CLASS_ALREADY_EXISTS`.
Pull Request: https://projects.blender.org/blender/blender/pulls/144231
With 5.0 we start requiring this extension for GL and VK.
All of our target hardware supports it with up to date
drivers.
Some old drivers were disabling this extension because of
buggy behavior. We simply drop support for them in 5.0.
This allows us to remove a lot of code and the last
shader create info override done at startup. This will
unlock more refactoring of the shader create info into
static classes to reduce binary size and other benefits.
## TODO:
- [x] Remove checks for ARB_shader_draw_parameters
- [x] Remove checks for ARB_clip_control
- [x] Check for the extension on startup for OpenGL
- [x] Check for the extension on startup for Vulkan
- [x] ~~Add user facing popup message about minimum
requirements not being met.~~ Done using the same
popup as old hardware.
Pull Request: https://projects.blender.org/blender/blender/pulls/142334
This PR will swap device local memory to system ram. It relies on
VK_EXT_external_memory and VK_EXT_pageable_device_local_memory
extensions to be supported by the system.
Most platforms support these extensions.
Pull Request: https://projects.blender.org/blender/blender/pulls/144422
* GHOST is an implementation detail, Blender environment variables should
have the BLENDER_ prefix.
* Don't put links in command line help output. We don't do this for other
arguments either, and it's trivial to search for this online.
* Make description more straightforward and line wrap.
* Use CLOG for logging.
Ref #143049
Pull Request: https://projects.blender.org/blender/blender/pulls/144349
The data constructed by this call remains in the 'C-alloc' realm, i.e.
it can be `MEM_dupallocN`'ed, and `MEM_freeN`'ed.
This is intended as a temporary API only, to facilitate transition to
full C++ handling of data in Blender. It's primary target is to allow
pseudo-POD types to use default values for their members. See e.g.
!134531.
Unlike !143827 and !138829, it does not change the current rule (`new`
must be paired with `delete`, and `alloc` must be paired with `free`).
Instead, it defines an explicit and temporary API to allow a very
limited form of construction to happen on C-allocated data, provided
that the type is default-constructible, and remains trivial after
construction.
### Notes
* The new API is purposely as restrictive as possible, trying to
only allow the current known needs (init with default member values).
This can easily be extended if needed.
* To try to stay as close as malloc/calloc behavior as possible, and
avoid the 'zero-initialization' gotcha, it does not use
value-initialization, but instead default-initialization on zero-
initialized memory.
_Ideally it would even not allow any user-defined default constructor,
but this does not seem simple to detect._
Pull Request: https://projects.blender.org/blender/blender/pulls/144141
This PR proposes to add an environment variable for forcing vsync to
be on or off. My primary use case was to disable vsync for forcing
viewport rendering performance tests not to be capped at the display
refresh rate when #142984 is used for removing animation frame rate
limits.
I initially added the environment variable "GHOST_VSYNC_OFF", but
found "GHOST_VSYNC=0/1" to be more easily understandable.
Usage:
- GHOST_VSYNC=0 => Vsync is forced off
- GHOST_VSYNC=1 => Vsync is forced on
Pull Request: https://projects.blender.org/blender/blender/pulls/143049
This fits better with the way normal and displacement maps are typically
combined. Previously there was a mixing of displaced normal and undisplaced
tangent, which was broken behavior.
Additionally, to undisplaced_N and undisplaced_tangent attributes must now
always be used to get undisplaced coordinates. The regular N and tangent
attributes now always include displacement.
Ref #142022
Pull Request: https://projects.blender.org/blender/blender/pulls/143109
Modify shader update so we simplify the graphs first to determine the
kernel features, then load the kernels, and only then update data on the
device. This avoids errors due to mismatched kernels and shaders.
Pull Request: https://projects.blender.org/blender/blender/pulls/144238
Resolve regression in [0] where setting the cursor wasn't checking
if a software cursor was used. This would then show the hardware &
software cursors.
[0]: 232b106e64
Swapchain handling of minimized windows wasn't correct. On some
platforms it still tried to create images with no surface.
This PR will discard swapchains of minimized windows, but still being
able to flush the render graph and free resources.
Pull Request: https://projects.blender.org/blender/blender/pulls/144189
Vulkan 1.0 render passes have been replaced by dynamic rendering in 1.2.
Blender Vulkan backend was implemented with dynamic rendering in mind.
All our supported platforms support dynamic rendering. Render pass support
was added to try to work around an issue with legacy drivers. However these
drivers also fail with render passes.
Using render passes had several limitations (blending and some workbench
features were not supported). As no GPU uses it and it is quite some code
to support it is better to remove it.
Pull Request: https://projects.blender.org/blender/blender/pulls/144149
This PR removes the deferred allocation of swapchain images. A lot of layers
don't support this and will crash when used.
Thanks to Jorn Visser to point out the potential issue.
Pull Request: https://projects.blender.org/blender/blender/pulls/143924
The Principled BSDF has a ton of inputs, and the previous SVM code just always
allocated stack space for all of them. This results in a ton of additional
NODE_VALUE_x SVM nodes, which slow down execution.
However, this is not really needed for two reasons:
- First, many inputs are only used consitionally. For example, if the
subsurface weight is zero, none of the other subsurface inputs are used.
- Many of the inputs have a "usual" value that they will have in most
materials, so if they happen to have that value we can just indicate that
by not allocating space for them.
This is a bit similar to the standard "pack the fixed value and provide
a stack offset if there's a link" pattern, except that the fixed value
is a constant in the code and we allocate a NODE_VALUE_x if a different
fixed value is used.
Therefore, this PR re-implements the parameter packing in a more efficient way:
- If we can determine that a component is disabled, all conditional inputs are
disconnected (to avoid generating upstream nodes).
- If we can determine that a component is disabled, we skip allocating all
conditional inputs on the stack.
- The inputs for which a reasonable "usual" value exists are changed to
respect that, and to only be allocated if they differ.
- param1 and param2 (which are fixed-value-packed as on all BSDF nodes) are
used to store IOR and roughness, which have a decent chance to be fixed
values.
- The parameter packing is more aggressive about using uchar4, which allows
to get rid of two SVM nodes while still storing the same inputs.
The result is a considerable speedup in scenes that make heavy use of the
Principled BSDF:
| Scene | CPU speedup | OptiX speedup |
| --- | --- | --- |
| attic | 5% | 9% |
| bistro | 5% | 8% |
| junkshop | 5% | 10% |
| monster | 3% | 4% |
| spring | 1% | 6% |
Pull Request: https://projects.blender.org/blender/blender/pulls/143910
Ever since the OptiX 8 update in Blender 4.5, the minimum GPU driver
requirements to use OptiX has increased to 535 or newer.
This commit update the minimum GPU driver requirement listed in the UI
to reflect this.
Pull Request: https://projects.blender.org/blender/blender/pulls/143917
This PR adds a new `fresnel_conductor_polarized` function, which calculates reflectance and phase shift (if requested) for both parallel and perpendicular polarized light. This is needed for applying thin film iridescence to conductors (see !141131).
For consistency, this PR also makes `fresnel_conductor` call `fresnel_conductor_polarized` instead of using a fast approximation of the Fresnel equations that is inaccurate at lower n and k values. This will change the output of some Metallic BSDF renders using Physical Conductor and prevent discrepancies when enabling thin film iridescence.
I didn't do any rigorous performance testing, but from timing the functions outside of Blender, `fresnel_conductor_polarized` is significantly slower than the approximation, between 1.5-3x depending on the compiler. This makes sense because it has three square roots and the approximation has none. In some informal tests with metallic_multiggx_physical.blend modified to have more spheres, the new renders took around 1-2% longer on both CPU and GPU.
There are some avoidable inefficiencies in this approach of just calling `fresnel_conductor_polarized`:
- one of the three square roots could be saved since `fresnel_conductor` never needs the phase shift and there are simplifications possible when only calculating the reflectance
- there are several unnecessary multiplications by 1.0 since `fresnel_conductor` uses relative IOR and `fresnel_conductor_polarized` doesn't, though those could get optimized out if inlined
Pull Request: https://projects.blender.org/blender/blender/pulls/143903
This replaces `stack_assign` with `stack_assign_if_linked`, which should save a few SVM nodes for constant parameters.
Running benchmarks (all scenes in the benchmark repo, 3 runs, median value for each) shows 1.0% improvement on CPU and 1.5% on OptiX. Not huge, but fairly (all between -0.2% and 3.0%).
Pull Request: https://projects.blender.org/blender/blender/pulls/143404
